Human Detection In Hours Of Darkness Using Gaussian Mixture Model Algorithm

Abstract

Video surveillance systems are very important in our everyday life. Video surveillance applications are used in airports, banks, offices and even our homes to remain us secure. Night vision is the ability to see in low light conditions. Surveillance video may not be seen clearly. Especially under the weak illumination conditions. The details of the image are very poor at night. Most of the previous work has focused on daytime Surveillance. This paper proposes a method of Human detection in hours of darkness (night time) using Gaussian Mixture Model (GMM) in real night environment employing an infrared radiation camera. Infrared video is taken as Input to perform Human detection at night. Then the video was processed using Gaussian mixture model algorithm, it was confirmed that by using this method accurate detection of human at night is possible.

Convolutional Neural Network-Based Human Detection in Nighttime Images Using Visible Light Camera Sensors

“…Because of these problems, single camera-based methods are being researched. In the case of single camera-based methods, there is research on using only IR cameras to perform human detection [ 14 , 15 , 16 , 31 ]. In [ 14 ], a Gaussian mixture model (GMM) was used on FIR image frames to find the background intensity distribution of each pixel.…”

“…In the case of single camera-based methods, there is research on using only IR cameras to perform human detection [ 14 , 15 , 16 , 31 ]. In [ 14 ], a Gaussian mixture model (GMM) was used on FIR image frames to find the background intensity distribution of each pixel. Then, when a new FIR image appeared, the probability of the background was calculated based on the background intensity distribution, and if this probability was low, it was determined to be the foreground and human detection was performed.…”

“…This is because the features needed for human detection sometimes do not exist, to the point where it is difficult to find human areas even with the human eye, and there is an outstanding amount of noise observable. Thus, there has been a large amount of research performed on human detection using far-infrared (FIR) cameras instead of the widely used visible light cameras [ 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 ]. Moreover, research has been performed on human detection by fusing visible light cameras and IR cameras to more accurately obtain features from nighttime images [ 27 ].…”

A Deep Learning Framework for Detection of Targets in Thermal Images to Improve Firefighting

“…Significant research has also been done in human detection in other dark/ low visibility environments utilizing a single visible camera and fusion of the generated RGB image with an IR data set. Single IR camera-based detection mechanisms have been presented in [14]- [17]. Most of these approaches use HOG-based feature extraction and a classifier using SVM or other ML-based techniques [16], [18].…”

“…Most of these approaches use HOG-based feature extraction and a classifier using SVM or other ML-based techniques [16], [18]. Paper [14] presents the use of a GMM system in human detection. Template matching techniques and thresholding techniques have also been reported in [16], [17].…”